Uso de relações de similaridade para tratamento de consistência e cobertura em sistemas de regras difusas / The use of similarity relations to deal with consistency and covering problems in systems of fuzzy rules

AUTOR(ES)

Isabela Neves. Drummond

DATA DE PUBLICAÇÃO

2003

RESUMO

This work presents the use of similarity relations to deal with inconsistency and covering problems that may occur in a knowledge base with rules of the type "If X is Ai then Y is Bi", where Ai and Bi are fuzzy sets. The work is focused on systems using implicative fuzzy rules based systems, i.e, in which the if-then operator is implemented by a truly implication operator, and a t-norm (the min operator) is employed to aggregate the output. The approach based on similarity consists of the use of similarity relations to enlarge the imprecision of the fuzzy sets employed in a given application. On this way, it is possible to solve inconsistencies, that occur when, for a valid input, the fired rules present conflict, and/or covering problems, that occur when, for a valid input, there are no rules whose premises address the input. Different strategies are defined: the global approach, where a similarity relation is applied to all the fuzzy terms, and the local approach, where a similarity relation is applied to all the terms, but the modified terms are used only for the inputs which fired conflicting rules. In this work, the constraints that a similarity relation must obey so that its application induces a minimum loss of information. A genetic algorithm is also employed in order to learn the best parameters that define the fuzzy terms and the similarity relations for a given application, aiming at an optimization of the system performance. This work also brings a comparative analysis between the performance of a conjunctive fuzzy controller of the Mamdani kind and the performance of an implicative controller (employing Rescher-Gaines, Godel and Goguen operators), on which the proposal approach was applied.

ASSUNTO(S)

sistemas difusos similarity relation inteligência artificial fuzzy sets fuzzy systems relação de similaridade conjuntos difusos artificial intelligence algoritmo genético sistema implicativo genetic algorithm implicative system

Documentos Relacionados

• The use of similarity relations to deal with consistency and covering problems in systems of fuzzy rules
• Sistemas baseados em regras fuzzy e aplicações
• Architecture and implementation of mechanisms to support business rules in information systems
• Arquitetura e implementação de mecanismos para suporte a regras de negócio em sistemas de informação
• Use of the correlation coefficient in agricultural sciences: problems, pitfalls and how to deal with them